An online lighting model estimation using neural networks for augmented reality in handheld devices

The level of realism in augmented reality applications is heavily dependent on the consistency of illumination between real objects and virtual objects. This paper presents a comprehensive methodology to model the real world lighting and synthesize it with virtual objects which are rendered. While there is a substantial body of knowledge on this aspect, the novel methodology suggested in this paper has its own advantages of not having to have any prior knowledge on the environment or any special hardware, which increases the usability of the system while making it possible to be utilized in online systems.